local-action-verification

使用 act 建立本地 GitHub Actions 驗證儲存庫,讓 Jules 能在推送前驗證 CI

npx skills add https://github.com/google-labs-code/jules-skills --skill local-action-verification

Local Action Verification with act

You are setting up a repository so that Jules (or any agent) can run GitHub Actions workflows locally using act to verify code changes pass CI before pushing.

What You're Setting Up

Two scripts and an agents.md section that enable local CI verification:

  1. install-act.sh — Installs act if missing (platform-aware, sudo fallback)
  2. run-act.sh — Runs act in the background with log polling to avoid agent timeouts
  3. AGENTS.md section — Instructions Jules reads to know how to use these scripts

Setup Steps

Step 1: Copy scripts to the repository

Copy the scripts/ directory from this skill into the target repository at scripts/act/:

Target structure:
scripts/act/
├── install-act.sh
└── run-act.sh

Make sure the scripts are executable:

chmod +x scripts/act/install-act.sh scripts/act/run-act.sh

Step 2: Add instructions to AGENTS.md

Append the following section to the repository's AGENTS.md file (create it if it doesn't exist). This is how Jules discovers the local verification capability:

## Local CI Verification

Before pushing code or opening a PR, verify changes pass CI locally using `act`.

### Prerequisites
- Docker must be running
- If `act` is not installed, run: `bash scripts/act/install-act.sh`

### How to Verify

1. Read `.github/workflows/` to find the CI workflow and identify the job ID
2. Run the verification script:
   ```bash
   bash scripts/act/run-act.sh "push -j <JOB_ID>"

With matrix: bash scripts/act/run-act.sh "push -j <JOB_ID> --matrix <KEY>:<VALUE>" 3. If the run fails, read the log output, fix the code, and re-run 4. After verification, clean up:

rm -f act_output.log
git checkout <any unintended file changes>

Configuration

  • Timeout: ACT_TIMEOUT=900 bash scripts/act/run-act.sh "..." (default: 600s)
  • Poll interval: ACT_POLL=15 bash scripts/act/run-act.sh "..." (default: 10s)
  • Custom image: pass -P ubuntu-latest=node:20-bookworm in the arguments for faster pulls

### Step 3: Update .gitignore

Append these entries to `.gitignore` if they don't already exist:

act artifacts

act_output.log .secrets


### Step 4: Print next steps for the user

Tell the user:
1. Docker must be installed and running on any machine (or Jules VM) where verification runs
2. `act` will be auto-installed on first use via `scripts/act/install-act.sh`
3. If workflows require secrets, create a `.secrets` file (KEY=VALUE format) — never commit it
4. Commit all generated files

## Troubleshooting

- **Docker not running**: `act` requires Docker. Ensure the Docker daemon is started.
- **Image pull slow**: First run downloads ~2GB+. Use `-P ubuntu-latest=node:20-bookworm` for faster pulls.
- **ARM64 issues**: On Apple Silicon, add `--container-architecture linux/amd64` to act arguments.
- **Secrets required**: Create a `.secrets` file and pass `--secret-file .secrets` in the act arguments.
- **Timeout**: Increase with `ACT_TIMEOUT=1200 bash scripts/act/run-act.sh "..."`.

## Resource References

- [Troubleshooting Guide](resources/troubleshooting.md) — Detailed solutions for common issues

來自 google-labs-code 的更多技能

typed-service-contracts
google-labs-code
使用「Spec and Handler」模式建構穩健、型別安全的 TypeScript 服務之架構標準。適用於建置 CLI、函式庫或複雜…
stitch::code-to-design
google-labs-code
將前端程式碼(Vite、React 等)轉換為 Stitch Design,透過鏈式提取靜態 HTML、設計系統及檔案上傳。**務必**使用此…
react-vite-dashboard
google-labs-code
將Stitch設計轉換為生產級React + Vite儀表板,搭配TanStack Query、來自DESIGN.md的可存取token,以及Web3就緒模式(ethers/viem)。
stitch::extract-design-md
google-labs-code
從前端原始碼中直接提取全面的設計系統(DESIGN.md)——支援 React、Vue、Svelte、Angular、純 HTML/CSS 或任何網頁框架。
remotion
google-labs-code
使用 Remotion 從 Stitch 應用程式設計中建立專業的逐步解說影片,包含流暢的轉場與文字疊加。從 Stitch 專案中擷取畫面,並將其編排成 Remotion 影片組合,支援縮放效果、淡入淡出轉場及情境文字疊加。支援模組化元件架構,包含 ScreenSlide 與 WalkthroughComposition 元件,以及互動式熱點與旁白整合等進階功能。產生畫面清單、下載...
ink
google-labs-code
用於 json-render 的 Ink 終端渲染器,可將 JSON 規格轉換為互動式終端 UI。適用於使用 @json-render/ink 從…構建終端 UI 時。
tdd-red-green-refactor
google-labs-code
此技能實作了一個結構化框架,用於AI輔助程式設計,確保每一行程式碼都可驗證、具型別且有意義。
automate-github-issues
google-labs-code
使用並行的 Jules 編碼代理設置自動化的 GitHub 問題分類與解決。